Future Video Reasoning on FutureBench Interp.
85.1AccuracyFUTURE-L1
Evaluation Results
| Method | Links | |
|---|---|---|
| FUTURE-L1Size=8B, Method=SFT+RL, Frames=322026.06 | 85.1 | |
| FUTURE-L1Size=8B, Method=SFT, Frames=322026.06 | 72.2 | |
| Video-CoESize=7B, Method=SFT+RL, Frames=322026.06 | 71.4 | |
| Video-o3Size=7B, Method=SFT+RL, Frames=322026.06 | 69.7 | |
| NEPSize=7B, Method=SFT+RL, Frames=322026.06 | 68.1 | |
| Qwen3-VLSize=30B-A3B, Method=Zero-Shot, Frames=322026.06 | 62.2 | |
| Text-Only SFTSize=8B, Method=SFT, Frames=322026.06 | 62 | |
| Video-R1Size=7B, Method=SFT+RL, Frames=322026.06 | 61.8 | |
| SwimBirdSize=8B, Method=SFT, Frames=322026.06 | 61.8 | |
| VideoAuto-R1Size=8B, Method=SFT+RL, Frames=322026.06 | 59.3 | |
| Qwen3-VL-InstructSize=8B, Method=Zero-Shot, Frames=322026.06 | 55.8 | |
| GPT-5Size=–, Method=Zero-Shot, Frames=322026.06 | 55.6 | |
| InternVL3Size=8B, Method=Zero-Shot, Frames=322026.06 | 54.4 | |
| Video-RFTSize=7B, Method=SFT+RL, Frames=322026.06 | 53.8 | |
| Qwen2.5-VL-InstructSize=72B, Method=Zero-Shot, Frames=322026.06 | 53.2 | |
| GPT-4oSize=–, Method=Zero-Shot, Frames=322026.06 | 51.6 | |
| Qwen2.5-VL-InstructSize=7B, Method=Zero-Shot, Frames=322026.06 | 50.7 | |
| MonetSize=7B, Method=SFT+RL, Frames=322026.06 | 49.7 | |
| GLM-4.1VSize=9B, Method=Zero-Shot, Frames=322026.06 | 47.3 | |
| LLaVA-NeXT-VideoSize=7B, Method=Zero-Shot, Frames=322026.06 | 44.4 | |
| MiMo-VLSize=7B, Method=Zero-Shot, Frames=322026.06 | 43.8 | |
| LVRSize=7B, Method=SFT+RL, Frames=322026.06 | 17.6 |